Natural Variation and Neuromechanical
Systems
Bradly Alicea
Michigan State University
East Lansing, MI. USA
Keywords: Neuromechanics, Biomimetic
Design, Human Variation, Stochastic
Systems, Human-Machine Interaction
ABSTRACT
Natural variation plays an important but
subtle and often ignored role in
neuromechanical systems. This is especially
important when designing for living or
hybrid systems which involve a biological or
self-assembling component. Accounting for
natural variation can be accomplished by
taking a population phenomics approach to
modeling and analyzing such systems. I will
advocate the position that noise in
neuromechanical systems is partially
represented by natural variation inherent in
user physiology. Furthermore, this noise can
be augmentative in systems that couple
physiological systems with technology.
There are several tools and approaches that
can be borrowed from computational
biology to characterize the populations of
users as they interact with the technology. In
addition to transplanted approaches, the
potential of natural variation can be
understood as having a range of effects on
both the individual's physiology and
function of the living/hybrid system over
time. Finally, accounting for natural
variation can be put to good use in humanmachine
system
design,
as
three
prescriptions for exploiting variation in
design are proposed.
INTRODUCTION
Variation can be a good thing in
engineered systems, especially when one or
more component is a living entity. In
neuromechanical systems and
more
generally in bioengineering applications,
variation plays an important role in adaptive
and maladaptive processes alike. Ultimately,
the existence of variation allows for
adaptation to new environments without the
recurring need for specialization (for
example, see Appendix 1, Number 1). This
is a lesson the engineering community is
currently learning by working with
biomimetic materials and structures [1, 2].
Biologically-inspired designs often include
materials or control strategies that are both
compliant (e.g. adaptive) and specialized to
a certain task [3, 4]. This broad survey and
conceptual framework will focus on how
this lesson can be transferred to the context
of neuromechanical technology design by
relating
variation
exhibited
during
performance and across individuals to the
unique optimality criteria for such systems.
Standardization and its discontents
The role of standardization in modern
engineering design has many advantages
from a mass production standpoint [5]. In
human-machine systems, however, mass
customization plays more of a role in
minimizing functional defects. Much as is
the case with internet technologies [6],
neuromechanical and bionic systems (for
definitions, see Appendix 1, Numbers 2 and
3) rely on user characteristics to achieve this
mass
customization.
However,
the
characteristics of interest in the latter type of
system are biological and physiological in
scope.
One interesting outcome that arises from
this is a general research approach that
distinguishes between fully synthetic
technologies (e.g. automobiles or laptop
computers) and living/hybrid technologies
(e.g. domesticated animals or prosthetic
limbs). A particular question can also be
posed: why is standardization generally a
good thing for fully synthetic machines (e.g.
automobiles), but bad for living industrial
systems (e.g. domesticated animals)? For
additional discussion, see Appendix 1,
Number 4. The short answer to this question
involves recognizing that what generally
holds true for synthetic technologies may
not hold true for living/hybrid ones.
Variation and Noise
The relationship between variation and
noise is also important in the design of both
synthetic and living/hybrid systems. Noise
arising from variation can serve a
constructive role [9]. One example of this is
stochastic resonance, which is based on a
combination of a patterned signal generated
from inside the system functionally coupled
with environmental noise. Stochastic
resonance can be defined as the alteration or
enhancement of a signal through the
addition of a stochastic, or noisy, component
[10]. In more technical terms, Mitaim and
Kosko [11] define stochastic resonance as
noise that enhances an external forcing
signal in a nonlinear dynamical system. This
is particularly important in neuromechanical
systems that involve vibrational or
oscillatory components. This could improve
the designs of a series of products ranging
from motorcycles to heart pacemakers.
Another example of beneficial noise,
which happens to be a particularly powerful
force in shaping living/hybrid systems, is the
intrinsic generation of randomness [12].
This type of noise involves randomness
expressed as part of processes internal to the
system. According to Wolfram [12], this
type of noise allows highly complex systems
to be constructed from rather simple rules,
thus demonstrating the constructive role of
noise and the need to consider living/hybrid
systems apart from purely synthetic ones.
In living/hybrid systems, these forms of
beneficial noise are closely tied to variation
observed both in real-time and over the
course of extended interaction. Specifically,
variation in living/hybrid systems involves
baseline
genetic
characteristics,
physiological processes, and regulatory
mechanisms. The expression of this
variation over the course of interaction is
triggered by these forms of beneficial noise.
The need for variation in living/hybrid
engineered systems
In both living/hybrid and synthetic
contexts, variation comes in two forms:
behavioral and structural. In synthetic
systems, behavioral and structural variation
not accounted for in the original design
often plays a deleterious role 1. Likewise,
structural and behavioral variation in
living/hybrid systems can play a deleterious
role. However, the inherent existence of this
variation can also allow the entity to adapt
given a range of environmental conditions.
In systems with a living/bionic component,
the ability to adapt allows for a diverse set
of responses. Two of these that will be
covered in this paper are 1) populationbased, genome-wide variation that sets the
baseline for performance in an individual,
and 2) expression mechanisms that augment
the differences characterized by these
baselines over the course of interaction
between living and non-living components.
INTRODUCTION
TO
NATURAL
VARIATION FOR ENGINEERING
Natural variation can be defined as the
genetic, hormonal, and morphological
diversity found among what will be referred
to herein as in situ populations. This is
especially important for understanding the
function of neuromechanical systems, which
can be defined as the biologically-salient
1
For examples from bridge structural mechanics, see
[7, 8].
study of how neuromuscular, environmental,
and movement variables are interrelated.
In situ populations
In situ populations can be defined in
contrast to statistical populations. In situ
populations represent groups belonged to by
the individual (e.g. familial or ethnic) being
sampled. While the full extent of variation in
each population not known, several
approaches based on existing techniques
used in organismal biology may allow us to
uncover the structure and relevance of this
variation.
There are three approaches I will
introduce in the next section that are, in the
context of this paper, specially tailored for
better understanding standing variation in
living/hybrid systems applied to in situ
populations. These are population genetics,
artificial selection, and gene expression and
transcriptional regulation models. However,
before we delve into methodology, the
specific role of natural variation on
living/hybrid must be clarified.
Understanding the role of biological
variation in hybrid/living systems involves
taking into consideration existing variation,
expression of variation across the lifespan,
and the by-products of evolutionary and
demographic processes. One aspect of this
variation involves biological differences
which may be amplified or repressed in
particular environments [13]. For example,
two individuals who carry unique genetic
and performance profiles to the task may
respond similarly in one environment but
much differently in another environment.
Furthermore, these differences may be
further mitigated and/or augmented by the
effects of aging or adaptation to the
environment in question.
On the other hand, switching between
two environments in a rapid and patterned
manner may encourage the expression of
previously hidden variation [14]. Such
results have been found in animal and
bacterial
models
in
response
to
environmental fluctuations, and may also
apply in a limited sense to human
neuromuscular systems. In a general sense,
environmental switching involves an
internal response to an oscillation. When
coupled with environmental settings where
external forces are applied to the body in
this
fashion,
complex physiological
responses may result that may only be
understood by looking at variation in the
structure and expression of the genome.
Population phenomics: an approach for
living/hybrid engineered systems.
In biology, phenomics [15] is the study
of phenotypes as they relate to
physiological,
cellular,
and
genetic
regulation.
This
is
an
important
consideration for design in a number of
technological systems. In this section, three
means to understand this diversity will be
covered: the use of population genetics
techniques, artificial selection experiments,
and gene expression and transcriptional
regulation models.
Population genetics techniques. Population
genetics tells us that traits are heritable [16].
Therefore, specific traits can be confined to
familial or ethnic lineages, which make up
varying proportions of any particular in situ
population. The relevance of populationlevel variation to neuromechanical systems
is that such variants set the baseline for
response
during
performance.
One
promising technique that can be used to
better understand how these types of variant
are distributed among in situ populations is
SNP genotypic [17]. Fortunately, emerging
technologies such as the HapChip 550K [18]
allow us to explore thousands of SNP
markers from across the genome in parallel.
Artificial selection experiments. Artificial
selection experiments with wearable devices
have the potential to trigger both adaptive
and maladaptive responses in cell
populations. These include epigenetic and
metabolic mechanisms which might be
selected for by using specially-designed
training regimens. This is in addition to the
normal
use
of
neuromechanical
technologies.
One approach that might be developed
further is the use of prosthetic devices which
change the force and surface properties that
the human user normally encounters during
movement [19]. One such device is a wand
with a large moment of inertia (I o) and a
forcing chamber at its end filled with
different types of materials. These materials
can be isotropic (such as water or sand),
electrorheological, or a matrix of custommade compliant, soft materials. When
swung or used to reach for objects in virtual
simulations, a degree of mismatch can occur
that decouples forces normally produced
during like movements under non-simulated
conditions. Changing the surface properties
in the forcing chamber and selectively using
the forcing chamber over time might either
mitigate or augment this effect in relation to
an evaluatory function described in the next
paragraph.
Artificial selection experiments are
typically used in an across-generation
context using model organisms [20].
However, artificial selection can also be
used to understand how variation affects
performance in behavioral contexts [21].
Artificial selection can be used to uncover
advantageous traits in an in situ population.
Given that all individuals in a given
population have a specific function for
evaluating performance2, selection for
performance can be applied by introducing
novel forces built into the neuromechanical
technological device.
Gene expression and transcriptional
regulation models. Of particular interest is
how the regulation of gene expression as a
functional response to external stimuli plays
a role in regulating performance. This
regulation is characterized by the output of a
physiological system. We can characterize
this variation in a theoretical manner, using
known information about the structure of the
genome and how environmental pressures
produce an adaptive response. The
techniques that can be borrowed from this
domain include gene regulatory networks
[22] and metabolic engineering models [23].
The take-home message is that physiological
regulation
forced
by
a
specific
neuromechanical technology can yield
differential results in different individuals.
Overall,
the
interaction
between
transcriptional
regulation,
metabolic
networks, and environmental stimuli can
trigger the additional expression of variation
in
situ
[14].
POTENTIAL
OF
NATURAL
VARIATION FOR SYSTEM DESIGN
The potential of natural variation in
system design can be characterized as the
relationship between robustness and
brittleness. Noise and variation are the most
important concepts that define these
variables. However, robustness and variation
can also be defined in terms of numbers and
relationships between real-world physics
2
a hypothetical specific function for evaluating
performance is analogous to fitness as defined in an
evolutionary context. Rather than being based on the
ability to produce greater or fewer offspring, the goal
of artificial selection in the context of this evaluatory
function is to favor specific adaptations and/or
maladaptations over a finite period of time.
(see Appendix 2, Sections 1 and 2,
respectively).
One way noise and variation play a role
in physiological responses and performance
outputs for specific performance settings is
by contributing to the nonlinear control
mechanisms needed to control complex
interacting systems. More specifically, the
interaction between environmental force
feedback and physiological response can
lead to effects such as dampening and
saturation, which result from the unfolding
of physiological processes over time as they
are selectively perturbed by variation
inherent in environmental feedback.
Examples
of
adaptation
and
maladaptation
Another way in which noise and
variation
affect
performance
in
neuromechanical systems is by mediating
how individuals adapt to a particular
interface or wearable device. When provided
with an environmental challenge, biological
mechanisms can rise to this challenge and
adapt immediately, adapt with some
training, or fail to adapt altogether. To better
characterize this phenomena, two variables
must be defined while specific examples of
adaptation and maladaptation must be given.
Specific example of adaptation: feedback
from interaction to physiology. During the
course of interaction between humans and
neuromechanical technologies, closed-loop
control is established between the
individual's physiology and control over the
technology at any one point in time. This is
why good ergonomic design usually requires
technological design to conform to natural
postures and do not introduce a lot of
extreme stresses, strains, and torsional
forces. This is hypothesized to lead to a
normal physiological response to technology
use.
Specific example of maladaptation:
dystonia. Dystonia is one example of how
mechanical technologies and individual
variation interacts in a deleterious manner
(see Figure 1). Dystonias of skeletal muscle
can be defined as muscle contractions that
cause twisting and repetitive movements and
overall abnormal function [24]. Focal
dystonia afflicts a specific set of muscles in
the hand and fingers. One classic example of
focal dystonia is the dysfunction of finger
muscles after the extended use of a stringed
instrument [25]. In this case, open-loop
control is responsible for the mismatch
between the mechanical system (in this case
a musical instrument) and the biological
system (in this case, muscles in a specific
part of the body).
APPLICATION OF VARIATION TO
NEUROMECHANICAL
SYSTEM
DESIGN
The variation found in a single individual
or group of individuals can be the basis for
nonlinear control strategies [26, 27]. It also
allows us to better understand how shortterm decrements in performance can lead to
more substantial long-term gains. Examples
might include the rate-limiting effects of
pharmaceutical agents, the dampening
effects of muscle and connective tissues, and
the feedback effects of physical training on
the regulation of gene expression. The
following sections explore and propose
ways in which this variation can be better
characterized
in
the
context
of
neuromechanical technology design.
Tool for design #1: variation and control
strategy curve.
One tool that can be generated from the
population phenomics approach to predict
performance in specific systems is the
variation and control strategy curve. This
curve is based on understanding how control
strategies employed during use of
neuromehcanical technologies relates to
expressed variation across a population.
We can use the Segway™ motor scooter
and Figure 1 as an example of how this
theoretical model works. For purposes of
expediency, we can characterize the
physiological metric as being the thigh-toshank ratio of the human lower extremity. In
the theoretical example, the thigh-to-shank
ratio across the population yields a Gaussian
probability density function (pdf). Within
the area of this pdf, other control strategies
are nested. It is of note that while other
control strategies are predicted to be nested
within the population-wide Gaussian, these
sub-functions need not be Gaussian
distributed themselves.
Figure 1. Graphical representation of the
variation and control strategy model.
In Figure 1, a sample distribution derived
from pseudodata is shown. As demonstrated
in Figure 1, subsets of the population are
representative of specific trait values. Of all
individuals characterized by this trait value
(a bin in standard histogram parlance), a
certain proportion of them will exhibit a
unique control strategy for a task give to the
entire population. In Figure 1, for the bin
characterizing a z-score of 0, 18% of
individuals
exhibit
a
second-order
dampening strategy. Likewise, 24% of
individuals exhibit a first-order dampening
strategy, and 58% a simple linear strategy.
In the case of a gyroscopicallycontrolled, electric powered scooter, a
plurality of individuals may use different
physiological control strategies when
challenged with a simple balance problem.
Artificially and
biologically-controlled
physical models [28, 29] suggest that a
simple linear control strategy is sufficient
for solving this problem using a passive
dynamic methods or a neural controller.
However, the above model suggests that
biologically-generated balancing using a
range of morphological parameters that
interface with a mechanical device can
produce multistability. Adding in ability to
produce muscle power might yield an even
greater number of control strategies used
across the population.
Tool for design #2: designing for faulttolerance using natural variation.
A better understanding of variation and
its
functional
consequences
during
interaction
with
neuromechanical
technologies may open the door to faulttolerant neuromechanical system design.
The standard approach to fault-tolerant
design in the field of computer networks
includes designing interactive systems with
several goals in mind [see 30, 31 for
inspiration of approach taken here]. Natural
variation can play a role in several of these
goals for neuromechanical systems, which
we will now review in a point-by-point
manner.
Redundancy. An understanding of how
physiological regulation responds to induced
loads allows us to distribute loads to
multiple anatomical points. This allows for
multiple modes of use across different
components of the population.
Isolation and containment of system
malfunction. A detailed understanding of
how genetic variants and biomechanical
function are interrelated allows for a way to
better contain maladaptive responses to
neuromechanical system use. This is
especially true over long periods of time
during which metabolic and other
physiological stresses induced by transitory
features of the mechanical system can turn
into deleterious conditions. Isolation and
containment at an early enough stage can
prevent these deleterious conditions. This
can also be accomplished in part by building
in safety mechanisms called reversion
modes.
Existence of safety mechanisms or
reversion modes. The ability to revert to
safe modes of operation in cases of
malfunction can be augmented by
accounting for natural variation in the design
of neuromechanical systems. One common
example of a reversion mode is the
Windows™ “safe mode”. The safe mode
can be deployed when the normal operating
parameters become corrupted. Likewise, a
neuromechanical system should allow for
the emergency shut-down of all nonessential
and
potentially
harmful
components of the interface once
maladaptation is considered probable by
means
of
physiological
indicators,
functional genomic data, and predictive
genomic markers.
EXPLOITING VARIATION IN DESIGN
There are three ways in which we might
exploit variation in design contexts:
genotyping individuals, designing for the
mapped physiological substrate, and
engaging specific traits with design. These
are not potential variables and relationships.
Rather, they are strategies for defining the
connection between measured/observed
physiological variation and the technological
and physical environment.
Individual variation in time-frequency
space. There is a need to redefine the
standard Gaussian definition of variation in
human physiology which will not be further
discussed here. In anthropometry, the sizes
of body segments and dimensions are
distributed according to a "bell curve"
model. While this is adequate for static
traits, such a model may be inadequate for
appreciating how these static traits
contribute to performance over the course of
technological interaction.
Figure 2. Top frame: relationship
between human and machine according
to traditional view (top frame) and
proposed view (bottom frame) 3.
One example of this stems from a
prediction
of
population
genetics:
polymorphisms which contribute to different
3
in Figure 2, the letter a) represents dystonia, a
specific maladaptive condition. Bottom frame: three
different humans interacting with the same machine.
Letter a) represents Human A being prone to
maladaptation, letter b) represents Human B having a
'normal' response to interaction with the machine, and
letter c) represents Human C having a dampened
response to maladaptive stimuli. The hypothetical
response observed in Human C is example of how a
high degree of robustness translates into system
design.
trait states are distributed according to a
power-law. In cases where there is diversity
at a particular locus, there tend to be a few
predominant variants and potentially many
variants at very low frequencies [32]. Rather
than a percentile model of understanding
variation, this new way of viewing variation
might be based on time-frequency
decomposition (for more information, see
Appendix 2, Section 3).
Defining a “mapped physiological
substrate”. There is also a need to
understand the relationship between
genotype and phenotype [33]. In the context
of the population phenomics approach, the
mapped physiological substrate can be
defined as what phenotypic precursors make
an individual more or less susceptible to
maladaptation. Getting at this question
allows us to better understand what we are
perturbing in the physiology with particular
technological designs. A more detailed
understanding of these new ergonomic
considerations may lead to the modification
of technology for specific components of the
population.
Integrating specific biological traits with
design. This point stems from defining a
mapped physiological substrate. Phenotypic
traits
with
complex
regulatory
underpinnings have greater opportunities for
adaptation during interaction. The dynamics
of a neuromechanical system in settings
such as this introduces a level of complexity
that goes well beyond the focal dystonia
example provided earlier. In such cases, a
nonlinear control system may be used to
tease out the complex effects of interaction
in various subsets of the in situ population
being sampled. For points on how this could
be applied to real-world systems, see
Appendix 2, Section 4.
CONCLUSIONS
In this paper, a number of concepts were
introduced that serve as a conceptual
framework for better understanding how
human variation plays a role in
neuromechanical system function and
ultimately effective design. These concepts
can be grouped into three general categories:
the role of noise and statistical distributions
of variation, methods for discovering
biological variation in so-called in situ
populations, and the integration of
mechanical and physiological function. Of
the three, the last of these points holds the
greatest potential for future development.
APPENDIX 1:
DEFINITIONS.
EXAMPLES
AND
Number 1: Specialization.
A simple example of this can be found
among bird's beaks. Among one group of
birds (species A), there is a single
morphology specialized to perform specific
tasks such as fighting or feeding. Members
of species A are said to be ecological
specialists. Among birds of species B, single
beak morphologies are also exhibited among
its members. However, in this species,
individual birds can exploit a wide range of
resources and perform a multitude of tasks.
Members of species B are said to be
ecological generalists. In species C, a range
of beak morphologies might exist. Members
of species C all perform the same sets of
tasks with divergent morphologies. It is this
latter type of variation seen in human traits
that might serve as an untapped resource
rather than a hindrance for purposes of
system design.
Number
2:
Neuromechanics,
Neuromechanical.
Neuromechanics can be defined as the
study of anatomical mechanics as it relates
to the nervous system. In many
neuromechanical systems, there is a tight
relationship
between
kinematics,
biochemical and muscle kinetics, and neural
control. Neuromechanical systems can be
better understood by examining examples
from humans, non-human animals, and
robots.
Number 3: Bionic systems.
Bionic systems are technological systems
that interface either closely or explicitly with
physiological
systems.
Biomedical
applications are the most high-profile
examples of bionic systems. However,
technologies that affect the nervous system
in a transformative manner can also be
considered to be bionic systems. Examples
include wearable or manipulable interfaces
that induce the effects of resistance exercise,
or technologies that induce specific
neuromuscular disorders such as dystonias.
Number 4: Differential standardization
considerations
in
synthetic
and
living/hybrid systems.
This definition is closely related to
specialization in that variation in different
types of technological systems is proposed
to have opposing effects. For example,
variation in synthetic systems may have
deleterious effects including production
defects which lead to malfunction. On the
other hand, variation in living systems
introduces additional levels of complexity to
product design.
In older products such as bicycles, the
recognition of variation as an assistive
component of the design process already
exists, albeit in limited form. In emerging
technologies such as
brain-machine
interfaces, individual variation may be both
an aid to customization and a barrier to
understanding uniform responses across
human populations.
APPENDIX 2: PARAMETER
ANALYTICAL DEFINITIONS
AND
Section 1: Robustness definition
Robustness can be defined as the ability
of the living/hybrid system to absorb
environmental
noise
challenges.
In
neuromechanical systems, this absorption is
a function of a technology designed to
minimize the resistance forces encountered
during movement. Robustness can be
measured using the parameter A4. In the case
of brittleness, A should be above 0.5 and in
extreme cases near 1.0.
Section 2: Brittleness definition
Brittleness must be defined as the
inability of the living/hybrid to respond to
an
environmental
noise
challenge.
Brittleness is typically a function of
technologies that create a large number of
resistance forces against some set of
muscles. Brittleness can also be measured
using the parameter A. In the case of
brittleness, however, the value of A should
be above 0.5 and in extreme cases near 1.0.
Section 3: Time-frequency approach to
variation
This provides inspiration for a new
approach to understanding biological
variation in the context of technological
systems. Specifically, a time-frequency
decomposition of combined physiological
and performance data is necessary. Such an
analysis is hypothetical at this point, but
may inspire future applications to specific
datasets. The proposed time-frequency
approach might involve the analysis of
genetic variants, the expression of genes
over time, and other variables that
4
the parameter A can be defined as the degree of
adaptability, which is the ratio of external forces
(measured in Newtons) against muscle power
(measured in Joules / meter · meter and representative
of internal forces).
characterize movement or muscle activity
over time [34].
Section 4: Approach to nonlinear control
and natural variation
One way this can be applied to design is
by mapping A) specific physiological traits
and B) suites of traits that define a
population to 1) specific trait states and 2)
their potential to be affected by feedback
from
environmental
forces.
These
relationships can be combinatorial so that
single and multiple trait relationships can be
related to either specific trait states such as
muscle hypertrophy or the robustness and
brittleness parameters (which define the
potential to be affected by feedback from
environmental forces).
REFERENCES
[1] Dillow, A.K. and Lowman, A.M.
Biomimetic Materials and Design. Reaction
Kinetics and Catalysis Letters, 80(2), 391397 (2003).
[2] Shin, H., Jo, S. and Mikos, A.G.
Biomimetic materials for tissue engineering.
Biomaterials, 24(24), 4353-4364 (2003).
[3] Choi, H.R., Jung, K., Ryew, S., Nam, JD., Jeon, J., Koo, J.C., and Tanie, K.
Biomimetic soft actuator: design, modeling,
control, and applications. IEEE/ASME
Transactions on Mechatronics, 10(5), 581593 (2005).
variation in manufacturing processes.
Quality Press, Milwaukee, WI. Page 23
(2005).
[6] Helms, M.M., Ahmadi, M., Jang, W.,
Jih, K., and Ettkin, L.P. Technologies in
support of mass customization strategy:
Exploring the linkages between e-commerce
and knowledge management. Computers in
Industry, 59(4), 351-353 (2008).
[7] Levy, M. and Salvadori, M. Why
Buildings Fall Down: how structures fail.
W.W. Norton, New York (1992). Section on
the Tacoma-Narrows bridge failure of 1940.
[8] Strogatz, S.H., Abrams, D.H., McRobie,
A., Eckhardt, B., and Ott, E. Theoretical
mechanics: Crowd synchrony on the
Millennium Bridge. Nature, 438, 43-44
(2005).
[9] Kosko, B. Noise. Viking Press. (2006)
Chapter 1.
[10] McNamara, B. and Wiesenfeld, K.
Theory of stochastic resonance. Physical
Review A, 39(9), 4854–4869 (1989).
[11] Mitaim, S. and Kosko, B. Adaptive
Stochastic Resonance. Proceedings of the
IEEE, 86(11), 2152-2183 (1998).
[12] Wolfram, S. A New Kind of Science.
Wolfram Media, Champaign, IL (2002).
[4] Okuno, R., Akazawa, K., and Yoshida,
M. Biomimetic myoelectric hand with
voluntary control of finger angle and
compliance. Frontiers of Medical and
Biological Engineering, 9(3), 199-210
(1999).
[13] Hoppeler, H., Klossner, S., and Fluck,
M. Gene Expression in Working Skeletal
Muscle. In "Hypoxia And The Circulation",
Advances in Experimental Medicine and
Biology series, Volume 618, Pgs. 245-254,
R.C. Roach, P.D. Wagner and P.H. Hackett
eds., Springer, Berlin (2008).
[5] Steiner, S.H. and Mackay, R.J. Statistical
Engineering: an algorithm for reducing
[14] Bergman A, Atzmon G, Ye K,
MacCarthy T and Barzilai N. Buffering
Mechanisms in Aging: A systems approach
towards uncovering the genetic component
of aging. PLoS Computational Biology,
3(8), e170 (2007).
[15] Freimer, N. and Sabatti, C. Human
genetics: variants in common diseases.
Nature, 445, 828-830 (2007).
[16] Hartl, D.L. and Clark, A.G. Principles
of Population Genetics. Sinauer Associates,
Sunderland, MA. (2006)
[17] Kim, S. and Misra, A. SNP
Genotyping: Technologies and Biomedical
Applications. Annual Review of Biomedical
Engineering, 9, 289-320 (2007).
[18] Loza, M.J., McCall, C.E., Li, L., Isaacs,
W.B., Xu, J., and Chang, B-L. Assembly of
Inflammation-Related Genes for PathwayFocused Genetic Analysis. PLoS One, 2(10),
e1035. doi:10.1371/journal.pone.0001035
(2007).
[19] Alicea, B. Performance Augmentation
in Hybrid Systems: techniques and
experiment.
arXiv
Repository,
arXiv:0810.4629 [q-bio.NC] (2008).
[20] Hill, W.G., and Caballero, A. Artificial
Selection Experiments. Annual Review of
Ecology and Systematics, 23, 287-310
(1992).
[21] Garland, T. Selection experiments: an
under-utilized tool in biomechanics and
organismal
biology.
In
"Vertebrate
biomechanics
and
evolution".
D.G.
Homberger, V.L. Bels, J.P. Gasc, and A.
Casinos, pgs. 23-56, BIOS Scientific,
Oxford, UK (2003).
[22] Longabaugh, W.J.R., Davidson, E.H.,
Bolouri, H. Computational representation of
developmental genetic regulatory networks.
Developmental Biology, 283, 1-16 (2005).
[23] Vemuri, G.N. and Aristidou, A.A.
Metabolic Engineering in the -omics Era:
Elucidating and Modulating Regulatory
Networks. Microbiology and Molecular
Biology Reviews, 69(2), 197–216 (2005).
[24]
NINDS
dystonia
fact
sheet.
http://www.ninds.nih.gov/disorders/dystonia
s/detail_dystonias.htm. National Institutes of
Neurological Disorders and Stroke (2009).
[25] Altenmuller, E.O. and Jabusch, H-C.
Focal dystonia:
diagnostic,
therapy,
rehabilitation. In "Human Haptic Perception:
Basics and Applications". Martin Grunwald
ed. Birkhauser, Basel, Switzerland (2008).
[26] Fazekas, C., Szederkenyi, G., and
Hangos, K.M. Linear and nonlinear control
of musculoskeletal systems. IASTED
International Conference on Modeling,
Identification, and Control. Lanzarote,
Spain, 219-224 (2006).
[27] Jagacinski, R.J. and Flach, J.M. Control
Theory for Humans: quantitative approaches
to modeling performance. Erlbaum Press,
Mahwah, NJ (2003).
[28] Pasemann, F. and Dieckmann, U.
Evolved neurocontrollers for pole-balancing.
In Biological and Artificial Computation:
From Neuroscience to Technology, pgs.
1279-1287, LNCS, Volume 1240, Springer,
Berlin (1997).
[29] Cluff, T. and Balasubramaniam, R.
Motor Learning Characterized by Changing
Lévy Distributions. PLoS ONE, 4(6), e5998
(2009).
[30] Koren, I. and Krishna, C.M. Faulttolerant Systems. Elsevier, Amsterdam
(2007). Chapter 4.
[31] Abd El-Barr, M. Design and analysis of
reliable and fault-tolerant computer systems.
Imperial Press, London (2007).
[32] Ewens, W.J. Mathematical Population
Genetics. Springer, Berlin (1979).
[33] Benfey, P.N. and Mitchell-Olds, T.
From Genotype to Phenotype: Systems
Biology Meets Natural Variation. Science,
320(5875), 495-497 (2008).
[34] Tsonis, A.A., Kumar, P., Elsner, J.B.,
Tsonis, P.A. Wavelet analysis of DNA
sequences. Physical Review E, 53(2), 18281834 (1996).